# Step 1: Import libraries
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from statsmodels.tsa.statespace.sarimax import SARIMAX
# Step 2: Load CSV
df = pd.read_csv('monthly_cv_with_macro.csv')
# Step 3: Filter for India
df_india = df[df['country'] == 'India'].copy()
# Step 4: Convert 'Year-Month' to datetime
df_india['Year-Month'] = pd.to_datetime(df_india['Year-Month'])
# Step 5: Aggregate duplicate months (sum volume, average macro)
df_india = df_india.groupby('Year-Month').agg({
'volume': 'sum',
'Diesel_USD_per_Liter': 'mean',
'Petrol_USD_per_Liter': 'mean',
'Monthly_GDP_USD': 'mean'
}).reset_index()
# Step 6: Sort by date
df_india = df_india.sort_values('Year-Month')
# Step 7: Set index and frequency
df_india.set_index('Year-Month', inplace=True)
df_india.index.freq = 'MS' # monthly start
# Step 8: Handle missing values
df_india['volume'] = df_india['volume'].interpolate() # linear interpolation for volume
for col in ['Diesel_USD_per_Liter', 'Petrol_USD_per_Liter', 'Monthly_GDP_USD']:
df_india[col] = df_india[col].ffill() # forward fill macro variables
# Step 9: Prepare target and exogenous variables
y = df_india['volume']
exog_cols = ['Diesel_USD_per_Liter', 'Petrol_USD_per_Liter', 'Monthly_GDP_USD']
exog = df_india[exog_cols]
# Step 10: Define SARIMAX model
model = SARIMAX(
y,
exog=exog,
order=(1,1,0), # ARIMA(p,d,q)
seasonal_order=(1,0,0,12), # Seasonal (P,D,Q,s)
enforce_stationarity=False,
enforce_invertibility=False
)
# Step 11: Fit SARIMAX
model_fit = model.fit(disp=False)
print(model_fit.summary())
# Step 12: Forecast 2026–2030 (60 months)
future_index = pd.date_range(start='2026-01-01', end='2030-12-01', freq='MS')
# Generate future exogenous variables with growth assumptions
last_12_months = df_india[exog_cols].iloc[-12:].values
future_exog = []
for i in range(5): # 5 years
for j in range(12): # 12 months
diesel = last_12_months[j,0] * (1.02**(i+1)) # 2% annual growth
petrol = last_12_months[j,1] * (1.02**(i+1))
gdp = last_12_months[j,2] * (1.03**(i+1)) # 3% GDP growth
future_exog.append([diesel, petrol, gdp])
future_exog = pd.DataFrame(future_exog, columns=exog_cols, index=future_index)
# Step 13: Get forecast
forecast = model_fit.get_forecast(steps=len(future_index), exog=future_exog)
forecast_mean = forecast.predicted_mean
forecast_ci = forecast.conf_int()
# Step 14: Plot historical + forecast
plt.figure(figsize=(14,6))
plt.plot(y.index, y.values, label='Historical Volume', color='blue')
plt.plot(forecast_mean.index, forecast_mean.values, label='Forecast 2026-2030', color='red')
plt.fill_between(forecast_ci.index, forecast_ci.iloc[:,0], forecast_ci.iloc[:,1], color='pink', alpha=0.3, label='Confidence Interval')
plt.title('Monthly CV Volume Forecast for India (SARIMAX)')
plt.xlabel('Year-Month')
plt.ylabel('CV Volume')
plt.legend()
plt.grid(True)
plt.show()
# Step 15: Save forecast to CSV
forecast_df = pd.DataFrame({
'Year-Month': future_index,
'Forecasted_Volume': forecast_mean.values
})
forecast_df.to_csv('India_monthly_cv_forecast_2026_2030.csv', index=False)
print("Forecast saved to 'India_monthly_cv_forecast_2026_2030.csv'")
# Step 16: Optional: print forecast
print(forecast_df)
